Estimation in Threshold Autoregressive Models with Adaptive LASSO: A Computational Evaluation

• Activity Number: 310
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Yiannis Kamarianakis and Maria van Schaijik*
• Companies: Arizona State University and Arizona State University
• Keywords: Threshold autoregressive models ; Lasso procedure ; Regime-switching models

Abstract

Threshold autoregressive models are parsimonious nonlinear specifications which have been found effective in capturing regime-switching dynamics in a wide variety of applications. This work proposes adaptive penalized estimation methods as an alternative to traditional stepwise procedures based on information criteria such as BIC. A battery of scenaria are simulated and the performance of LASSO-type (adaptive-LASSO, ordered LASSO) estimators is compared to the outcomes of stepwise model building procedures based on AIC, BIC and HQ. Finally, the forecasting performance of the alternative methods is tested on time series of road traffic volumes from an urban network.